Recent debates surrounding the inclusion of caste in 2011

Transcription

1 SPECIAL ARTICLE Caste in 21st Century India: Competing Narratives Sonalde Desai, Amaresh Dubey Recent debates regarding inclusion of caste in 2011 Census have raised questions about whether caste still matters in modern India. Ethnographic studies of the mid-20th century identified a variety of dimensions along which caste differentiation occurs. At the same time, whether this differentiation translates into hierarchy remains a contentious issue as does the persistence of caste, given the economic changes of the past two decades. Using data from a nationally representative survey of 41,554 households conducted in 2005, this paper examines the relationship between social background and different dimensions of well-being. The results suggest continued persistence of caste disparities in education, income and social networks. These results are based on the India Human Development Survey, This survey was jointly organised by researchers at the University of Maryland and the National Council of Applied Economic Research. The data collection was funded by grants R01HD and R01HD from the National Institutes of Health to the University of Maryland. Part of the sample represents a resurvey of households initially interviewed in the course of the Human Development Profile of India conducted by NCAER. Sonalde Desai (Sonalde.desai@gmail.com) is with the University of Maryland and Amaresh Dubey (amaresh.dubey@gmail.com) is with the Jawaharlal Nehru University, New Delhi. 40 Recent debates surrounding the inclusion of caste in 2011 Census have served to crystallise the competing narratives about social stratification in Indian society. One side holds that historical fault lines along caste, tribe and religious lines persist and may well have been aggravated in modern India; the other side asserts that, while caste remains an important dimension of modern social life, its relevance is mostly limited to selection of marriage partners and has little importance in shaping material inequalities. This brief description masks considerable complexities within each narrative. The narrative of inequality ranges from a vision of status inequality in which the ideology of purity and pollution is so ingrained that the subordinated groups willingly become part of a status hierarchy and are implicated in their own subjugation (Gerth and Mills 1946; Dumont 1980) to the vision in which exploitation of marginalised groups key to the success of the dominant group (Omvedt 2006). Similarly, the narrative which focuses on irrelevance of caste ranges from a historio graphy focusing on the role of the colonial administration in solidifying caste boundaries that were fluid until the arrival of East India Company (Dirks 2001) to one in which modern economy has served to increasingly blur the boundaries of caste-based occupational segregation, and thereby, narrowing inequalities (Gupta 2000; Kapur et al 2010). Empirical examination of these competing claims forms the core of this paper. Using data from India Human Development Survey of (IHDS), this paper examines each of these claims. The IHDS is a nationally representative survey of 41,554 households organised by researchers from the University of Maryland and the National Council of Applied Economic Research. It is a multitopic multi-purpose survey containing information about a variety of dimensions of social and economic well-being of the households. These data are in public domain 1 and at an all-india level, poverty, education, household structure and employment levels recorded in this survey are comparable to those recorded by Census and the National Sample Survey albeit with some exceptions associated with survey design (Desai et al 2010). 1 Competing Narratives While caste affiliations remain ubiquitous in modern India with surnames, marriage arrangements, dress and food habits often characterising caste distinctions, the extent to which caste defines the fundamental structure of social stratification in India has become a subject of contentious debates. Four perspectives deserve particular attention: 1.1 Caste as a Status Hierarchy One of the reasons caste has excited sociological imagination is because it is seen as a representation of pure status, based on march 12, 2011 vol xlvi no 11 EPW Economic & Political Weekly

2 religious and ideological grounds (Milner 1994; Dumont 1980; Weber 1958) with class inequalities being epiphenomenal to caste. This disjunction between the sacred and the profane gives the Indian caste system a sociological character that sets it apart from other forms of social inequality based on material resources. Weber lays out the essential characterisation of the caste system as opposed to affinity groups that has undergirded much of the sociological discourse on caste: A status segregation grown into caste differs in its structure from a mere `ethnic segregation; the caste structure transforms the horizontal and unconnected coexistences of ethnically segregated groups into a vertical social system of superordination and subordination.ethnic coexistences condition a mutual repulsion and disdain but allow each ethnic community to consider its own honor as the highest one; the caste structure brings out a social subordination and an acknowledgement of more honor in favour of privileged castes and status groups (Gerth and Mills 1946: 189). While Weber largely relied on writings by colonial bureaucrats in the Indian Civil Services (acting as amateur anthropologists) for data on Indian society, anthropological villages studies of the 20th century by Indian as well as western scholars provided a foundation for Louis Dumont s work (Dumont 1980). With the publication of Homo Hierarchicus in 1966, Dumont presented a canonical formulation that has framed the conversation about caste over the past four decades and provided a rationale for status hierarchy. In emphasising the ideological over the material his formulation has much in common with his predecessors (Weber 1958) and successors (Milner 1994). This narrative of caste has excited tremendous passions from diverse groups with wide-ranging critiques (Appadurai 1986; Gupta 2000; Kolenda 1973; Marriott 1969; Srinivas 1996; Berreman 1991). Status theories of caste hierarchies have a tendency to focus on ageless and timeless India as represented in vedic traditions, partly because they draw upon the religious foundations of caste. This focus often ignores modern India, particularly urban India, in which concepts like purity and pollution are difficult to implement in day-to-day life. As Andre Beteille (Dumont and Beteille 1987) remarks in an acerbic exchange with Dumont, Dumont s lack of ease with modern India is writ large in his work, although it does not shine as brightly as his enthusiasm for traditional India, which is partly an India of his own construction.. Modern India, in Dumont s construction, is not made of whole cloth, it is a thing of shreds and patches (Beteille: 675 in Dumont and Beteille 1987). While there seems to be a general agreement regarding social differentiation between castes based on visibly recognisable symbols, including rituals, dress, tonsorial styles and a host of other behavioural markers, whether this differentiation translates into social hierarchies in modern India is far from clear (Gupta 2000) and some intriguing studies have documented declining salience of caste over time even in rituals and food habits (Mayer 1997; Kapur et al 2010). 1.2 Caste as a System of Exclusion and Exploitation Research on caste as a system of exclusion and exploitation stands in sharp contrast to the Weberian focus on status hierarchies, in which the subordinate groups accept their low status. Given the occupational underpinnings of the varna system and the linkages SPECIAL ARTICLE between occupation and income, it is not difficult to see caste as a system of material inequality. Research on caste inequalities and their material basis has a long history (Gough 1981; Mendelsohn and Vicziany 1998) going back to some of the earliest research on jajmani systems (Wiser 1979). Even in modern India, scheduled castes (SCs) continue to dominate the ranks of the sweepers (safai karmacharis). SCs form nearly 60% of the sweepers in central government compared to only 18% of other Class D workers (GoI 2006). An interesting analysis of the Indian Administrative Services (IAS), shows that as late as 1985, about 37% of the IAS officers selfidentified as being brahmin (Goyal 1989), a disproportionately large number since brahmins form only about 5% of the population. Moreover, since a substantial proportion of the applicants do not provide caste information, this number is undoubtedly an underestimate. Consequently, it is not surprising that many studies have found a strong link between caste and economic status (Deshpande 2000; Thorat and Newman 2009) postulated to be a function of exclusion from access to productive resources such as land and education as well as discrimination in the labour market. Two aspects of caste inequalities deserve attention: inequality of opportunity and inequality of outcome. Centuries of castebased social organisation have left a legacy of inequality in access to land, education, business ownership and occupation. These processes lead to unequal access to productive resources and thereby lead to material disadvantages. However, caste-based inequalities are not simply limited to inequality in opportunity. It has also been argued that even highly qualified members of lower caste face social and economic discrimination resulting in inequality of outcomes (Thorat and Newman 2009). The distinction between inequality of opportunity and that of outcome is not straightforward; inequality of outcome in one generation may lead to inequality of opportunity in the next but this distinction remains important from a public policy perspective. 1.3 Caste in Transformation However, these linkages between caste, occupation and income are not accepted uncritically. It has sometimes been argued, the relationship between caste and occupation has been much misrepresented. It is doubtful that there was at any time a complete correspondence between the two. At any rate, even before independence many castes, and probably most, had more than half their working members in occupations other than those specifically associated with their caste (Beteille 1992: 40). In an independent India the link between caste and occupation has weakened considerably. The jajmani system has all but vanished, allowing for market-based pricing for services rendered by the workers (Commander 1983). Additionally a variety of forces have disrupted the link between caste and occupation. Land reforms transferred landownership to many former sharecroppers, most of whom belong to the middle castes (Dantwala 1950); declining incomes of artisans and influx of mass-produced goods have led to declining caste-based occupations among potters, weavers and other artisans who must now rely on manual labour for subsistence (Bayly 1999); and increased requirements Economic & Political Weekly EPW march 12, 2011 vol xlvi no 11 41

3 SPECIAL ARTICLE for education among modern professions have led to influx of people from a variety of castes into modern occupations (Sharma 1999). All of these trends would suggest that the link between caste and economic status in modern India is marginal at best. In an analysis of the numerically preponderant dominant castes in south India, noted anthropologist M A Srinivas found that certain peasant castes enjoy numerical superiority as well as political and economic power, although they remain middle castes by the varna schema (Srinivas 1987). Politics of affirmative action has further strengthened the power of lower castes with reservations in government jobs and higher education (Beteille 1992). Recent studies further document the dilution of the role of caste in shaping economic well-being and suggest that migration, expansion of dalits in non-traditional occupations and changes in agriculture combine to improve the relative position of dalits in recent years (Kapur et al 2010). 1.4 Caste as a Social Construction The literature on social and material dimensions of caste stands in marked contrast to the literature that emphasises the role of the colonial power in constructing caste as an enumerative community and thereby solidifying hitherto fluid identities (Das 2003; Dirks 2001). An interesting aspect of this critique includes debates about whether brahmins are at the pinnacle of the caste system or whether they were placed there by colonial imagination either because of a fascination with the exotic or as a way of creating a social order that could be exploited in colonial governance (Dirks 2001). This critique suggests that the presumed brahmin superiority is a creation of an orientalist discourse that privileged brahmins over Kshatriyas (Dirks 2001; Raheja 1988) and ignored the claims to ideological superiority from lower castes (Khare 1991; Chatterjee 1992; Das 1982). By focusing on alternate sources of legitimacy and hierarchy, these studies question the notion that the brahminical values are internalised by all subordinate castes and that caste hierarchies are accepted without interrogation. A view that sees caste as a construction of the imagination of the armchair anthropologists of colonial administration has interesting implications for the modern discourse on caste since it suggests that it is the very attention to caste as an organising principle of Indian society that leads to caste mobilisation and polarisation and without this continued attention, caste distinctions would wither away. 2 Debates without Data This brief discussion of the debates surrounding caste hierarchies in India and the competing theories about the nature and magnitude of upper caste control over ideological and material resources clearly suggests a need for empirical examination. However, many of these debates have been conducted in a sterile environment devoid of empirical substantiation using recent data. While the National Sample Surveys can provide some data on material well-being, they are of limited utility in studying social exclusion or subordination. Even for research on material inequalities, their range is somewhat limited. These paper tries to fill some of these gaps by focusing social and economic behaviours and outcomes of different social groups. 42 Several caveats about the IHDS and the analysis presented in this paper are important to note. The IHDS is the only nationally representative survey containing data on income, education and indicators of social interaction. Thus, it provides a unique opportunity to examine some of the debates surrounding caste. However, income is notoriously difficult to measure and in spite of considerable care taken in designing a questionnaire that enumerates over 50 sources of income, measurement errors cannot be ruled out (for potential sources of error see Desai et al 2010). Moreover, the analysis presented here is presented at a pan-india level; regional patterns in caste inequalities are likely to be quite distinct given different patterns of caste mobilisation in different areas. Finally, for simplicity of presentation, the analysis is restricted to Hindus and adivasis who are either Hindu or follow a tribal religion. Muslims, Christians, Jains, Sikhs and other religious groups are excluded. 2 The analysis focuses on two types of outcomes. Dichotomous variables such as ability to read or participation in a political meeting are analysed using logistic regression; continuous variables such as household consumption expenditure or years of education are analysed with multiple regression. In case of years of education, social networks and number of organisational memberships, we focus on regressing the mean of the distribution on variables of interest. Where the dependent variables are highly skewed such as the annual incomes of men and consumption expenditure, we use median regression where median rather than the mean of the distribution is regressed on variables of interest. Each regression controls for place of residence and state of residence. As appropriate, controls for own or parents education are also included. Tables 1-4 report marginal differences between different caste groups by setting values of all other variables at their mean values. The variables included in this analysis are shown in Appendix 1 (p 49); the analytical model, sample and included variables are described in Appendix 2 (p 49). The dependent variables of interest are listed below: 2.1 Civic and Political Participation Number of Organisational Memberships: The survey asked respondents whether any household members belonged any of the following organisations Mahila mandal; youth organisations focusing on sports or reading; unions or business organisations; self-help groups; credit or saving associations; development group or NGO; and, pani panchayat or other cooperatives. A count of the memberships in these organisations is used as dependent variable using in a regression estimated with OLS. Whether any household members participated in a political meeting called by the panchayat or nagarpalika ward in the year preceding survey: This is a dichotomous variable analysed with logistic regression. 2.2 Inequality of Opportunity Landownership: Whether the household owns and cultivates any land. This is a dichotomous variable analysed with logistic regression. Lack of access to land is an important determinant of rural poverty. march 12, 2011 vol xlvi no 11 EPW Economic & Political Weekly

4 Years of Education: Completed years of education for men aged analysed using OLS. Education forms one of basic aspect of labour market disadvantage. We focus on education for men aged since the subsequent analyses use this as a control variable for income, however, results not reported here show similar patterns of educational inequalities for both men and women. Household Consumption Expenditure Adjusted for Household Size: Household consumption expenditure determines the access to food, healthcare and education. Since household structure is an important determinant of access to resources, it is necessary to adjust for household size. However, instead of using per capita consumption expenditure, following recent suggestions in the literature on poverty and inequality (OECD 2008; Citro and Michael 1995), we take into account the economies of scale and divide total expenditure by the square root of the household size. Given the skewness of the distribution of expenditure and income, all income and expenditure outcomes are examined using median regressions where median of the distribution is regressed on variables of interest rather than the mean (Koenker and Bassett 1978). Social Networks: Social networks form an important avenue of access to job-related information as well as access to credit. The IHDS asked whether household members knew anyone who worked for the government, in schools or in the medical field. These three sets of contacts are summed and regressed on caste and other background variables using OLS. 2.3 Inequality of Outcomes Annual Earnings: The IHDS collected detailed data on over 50 sources of income. It also collected information on work participation for each household member including his or her intensity of work participation. Using these two sources of information, annual earnings for each individual are calculated using their pro-rata share of the income given their intensity of participation. For wage and salary income, each individual s earnings are attributed to their own share of the household income. For farm or business income, where many household members participate, their share of income is based on their intensity of participation. The annual earnings are regressed on caste, education and other background variables using median regressions. This analysis focuses on males aged in order to focus only on prime working ages and exclude biases introduced by differential rates of work participation. Annual Wage and Salary Income: Market discrimination based on caste is most likely to occur in wage and salary employment where employers directly control the kind of work opportunities available to individuals as well as their remuneration. Thus, this paper regresses the median of wage and salary income on caste, education and other background characteristics. Ability to Read a Short Paragraph: The IHDS administered short reading tests developed by PRATHAM (Pratham 2005) to SPECIAL ARTICLE children aged For this analysis, ability to read a short paragraph is used as a dependent variable in a logistic regression. The analysis controls for child s family background by controlling for highest education attained by any adults in the household, log of per capita household consumption expenditure, whether the household engages in farming, place and state of residence as well as child s sex. 3 Empirical Results The following sections present results from multivariate analyses. In each case, the results present marginal differences between various caste groups, holding other control variables at their mean value. The significance tests report whether the differences between the forward castes (the omitted category) and the index caste group are statistically significant. 3.1 Caste, Social Distance and Relations of Subordination Dumont s canonical formulation relies on willing acceptance on the part of lower castes to accept their subordinate status. Operationalisation of status is a challenging task, particularly in a society in which external markers of social status are changing rapidly. Much of the earlier literature has relied on manner of clothing, engagement in ritualistic behaviours and food habits. However, as a recent study perceptively points out (Kapur et al 2010), interpreting underlying social relations from observed behaviour is a difficult task. Kapur et al (2010) find declining distance between castes at wedding celebrations, paradoxically coupled with lower participation of lower castes in upper caste weddings. They argue that, poverty and dependence might explain why more dalits attended non-dalit weddings in 1990, even though separate seating was more a norm then. By 2007, though such humiliation had become rare, fewer dalits were keen on attending non-dalit weddings. It is a mark of dalits new-found independence both from upper castes and the food in their feasts (ibid 2010: 46). Departing from a focus on social intercourse, a number of other studies have tended to focus on political sphere to suggest irrelevance on caste in modern India. While few scholars claim a demise of caste, there is an increasing tendency to view caste as a form of ethnicity in which castes compete with each other for power and proudly brandish their own narratives of origin, with even the lowest castes claiming a place in the national history characterised by valour and accomplishment (Narayan 2004). Dissenting from an orientalist viewpoint, which saw the subjugated as being acquiescent in their own subjugation, recent research has tended to focus on the agency and power of the oppressed (Khare 1991). This dissent has been bolstered by political developments in modern India. As the era of grand ideology came to an end in India as elsewhere, regional political parties have emerged as power brokers (Brass 1990; Frankel et al 2000). Many of these parties rely on numerically large middle and lower castes for support; consequently, there is a perception that the varna structure has transformed itself into an ethnic jati structure, where small endogamous jatis compete with each other for social and political power. The politics of affirmative action sharpened this competition and led to increased attempts to Economic & Political Weekly EPW march 12, 2011 vol xlvi no 11 43

5 SPECIAL ARTICLE capture the support of the state. Consequently, it is argued that the middle and lower castes have staked out a claim to economic and political power that is on par with, if not greater than, the brahmins and other upper castes (Gupta 2005). Table 1: Predicted Probability of Participation in Formal Organisations and Attendance at a Political Meeting The IHDS provides direct measures of civic and political participation. The IHDS asked respondents about participation of any household members in a variety of civic organisations. It also asked whether anyone had attended a public meeting called by a panchayat or a ward committee in the preceding year. These two indicators allow us to study the extent of civic and political participation by different households. Results from this analysis are presented in Table 1. Although more educated and wealthier households are more likely to participate in civic and political activities, controlling for these factors, we see greater participation on the part of adivasis and dalits than other castes. After adjusting for regional, educational and landownership differences by setting these variables at their mean values, average value for number of civic association memberships is 0.36 for forward castes, while it is 0.47 for dalit households, and this difference is significant at 0.01 level. Similarly, holding other factors constant, 26% of the forward caste households attended a political meeting in the preceding 12 months compared to 28% dalit households. This suggests that dalits in the 21st century are far more politically active than the forward castes. While adivasis are somewhat less active than dalits, they also seem more politically engaged than the forward castes. A variety of factors may underlie this phenomenon. Social groups that feel economically oppressed may have more reasons to become politically active. Government programmes and policies may promote greater incentives for collective action by setting aside funds for dalit and adivasi self-help groups. But regardless of the reason, the results presented above paint a picture of a civic culture in which dalits, and to a lesser extent, adivasis, are taking a lead. Instead of suggesting a sense of ingrained subordination or sense of disenfranchisement, this observation points to active political participation on the part of lower castes that bodes well for long-term future of the Indian civil society. 3.2 Caste and Inequality Does greater participation by lower castes in political discourse suggest vanishing importance of caste? This would be an attractive argument for many reasons. The narrative of ethnicity instead of hierarchy has gained ground in recent decades, possibly because modern Indian intelligentsia continues to define caste through the eyes of a religious discourse that they find increasingly irrelevant. Moreover, urban India has enthusiastically embraced a 44 Membership in No of Civic Organisations Attended a Political Meeting in 12 Months Forward castes (ex brahmin) Brahmin OBCs Dalit 0.47 ** 0.28 ** Adivasi 0.43 ** 0.28 ** <= p 0.01 in comparison with forward castes (excluding brahmin). * <= p 0.05 in comparison with forward castes (excluding brahmin). Predicted values from the regression holding all control variables at their mean value. global discourse of equality where anyone who speaks against equality in public is bound to lose his audience (Beteille 1991). As Chris Fuller (1997:13) in the introduction to a volume titled Caste Today notes, The elimination of hierarchical values from legitimate public discourse accounts, for the claim that there is no caste left. That caste hierarchy can no longer be legitimately defended in public has itself contributed to the emergence of a more or less acceptable public discourse about status coded as cultural difference. Because people cannot openly speak of castes as unequal, they describe them as different. However, not all outcomes are like food choices, a matter of preference. In this paper we focus on two sets of outcomes, the first set describes the inequality of opportunity, or what is often called pre-market inequality (Table 2). The second set describes the inequality of outcome (Table 3). Table 2 shows statistically significant caste disparities in each marker of opportunity structure with dalits and adivasis at the bottom and Other Backward Classes (OBCs) in the middle. When compared to forward castes, dalits and adivasis are less likely to own land, have fewer years of education, have lower household size adjusted household consumption expenditure and have fewer important social connections. For example, holding place of residence constant, forward caste men aged completed 8.18 years of education on average, while OBCs complete only 6.68 years, dalits 5.23 years and adivasis 4.39 years. Similarly, annual household expenditure adjusted for family size by dividing with square root of the household size, is Rs 19,857 for forward castes compared to 17,961 for OBCs and 16,832 for dalits and 16,062 for adivasis. Forward castes also have more social network connections than OBCs, dalits and adivasis. All of these differences are statistically significant at 0.01 level and control for relevant education and other background variables. The included variables are described in Appendix 2. Table 3 examines the caste differences in outcomes controlling for education, place of residence and state of residence. Like Table 2, once again we see statistically significant differences between forward caste and dalits and adivasis in each outcome. Table 2: Inequality of Opportunity Controlling for Residence and Family Background Landownership Years of Educated Annual Consumption No of Social Rural Households Men Expenditure Adjusted Network for Household Size Connections Forward Castes , Brahmin 0.66 ** ** 20,320 * 1.20 ** OBCs 0.64 ** 6.68 ** 17,961 ** 1.03 ** Dalit 0.44 ** 5.23 ** 16,832 ** 0.98 ** Adivasi 0.63 ** 4.39 ** 16,062 ** 0.93 ** ** <= p 0.01 in comparison with forward castes (excluding brahmin). * <= p 0.05 in comparison with forward castes (excluding brahmin). Predicted values from the regression holding all control variables at their mean value. Table 3: Inequality of Outcomes Controlling for Education and Family Background Annual income Wage+ Annual Wage and Salary Ability to Read Farm+ Business Income for Males Simple Paragraphs (Males 25-49) in Wage Labour Children Ages 8-11 Forward castes 22,057 23, Brahmin 21,350 27,712 ** 0.69 * OBCs 19,934 ** 21,312 ** 0.58 Dalit 19,820 ** 21,712 ** 0.48 ** Adivasi 18,464 ** 20,812 ** 0.51 ** ** <= p 0.01 in comparison with forward castes (excluding brahmin). * <= p 0.05 in comparison with forward castes (excluding brahmin). Predicted values from the regression holding all control variables at their mean value. march 12, 2011 vol xlvi no 11 EPW Economic & Political Weekly

6 The OBCs are also below the forward castes in both measures of income income from all sources as well as just wage and salary income. However, the difference between forward caste and OBCs is not statistically significant for children s reading skills. All of the regressions in Table 3 control for pre-market differences between different caste groups. Hence, lower earnings of dalits and adivasis compared to forward castes are not just due to differences in education between the two groups that were recorded in Table 2 and suggest a possibility of market discrimination rather than simple disparities in access to education. Lower skill attainment of children aged 8-11 is particularly worrisome since the regressions control for family s standard of living as well as education of adults in the household. So these results suggest that even for children from similar socio-economic backgrounds, something about school environment results in lower levels of skill acquisition on the part of dalit and adivasi children, resulting in a long-term cycle of disadvantage. 3.3 Persistence of Caste Inequalities: The initial discussion in this paper noted the widespread belief that caste disparities are characteristics of a bygone era and have no place in modern India. Since the IHDS collected data in , it is not possible to use this survey to examine changes in caste disparities over time. However, since the sample contains urban as well as rural areas, it is possible to compare changes in caste disparities across different social and economic conditions. Table 4 divides the sample into metropolitan cities, other urban areas, developed villages blessed with higher levels of infrastructure such as roads and access to banks and other facilities and less developed villages. Since the metropolitan sample is relatively small only 3,337 households out of the sample of 41,554, the results need to be treated with caution. Nonetheless, a Table 4: Predicted Urban/Rural Differences in Caste Inequalities Metro Cities Other Urban Developed Villages Less Developed Villages Consumption expenditure adjusted for household size Forward castes 31,191 27,122 18,280 15,185 Brahmin 31,987 26,524 17,929 16,936 ** OBCs 28,930 ** 24,224 ** 16,381 ** 14,259 * Dalit 28,979 * 22,668 ** 15,281 ** 12,986 Adivasi 31,025 22,390 ** 14,568 ** 12,011 ** Annual income from wage, cultivation and business for males Forward castes 52,102 37,532 16,558 12,964 Brahmin 62,302 ** 36,757 18,346 * 12,440 OBCs 49,502 36,757 ** 15,476 * 11,900 Dalit 51,093 31,276 ** 14,142 ** 12,234 Adivasi NA 28,559 ** 14,053 ** 9,933 ** Ability to read children ages Forward castes Brahmin 0.76 x x 0.66 OBCs Dalit ** 0.45 ** 0.48 ** Adivasi NA 0.52 x ** x <= p.10 in comparison with forward castes (excluding brahmin). ** <= p 0.01 in comparison with forward castes (excluding brahmin). * <= p 0.05 in comparison with forward castes (excluding brahmin). Predicted values from the regression holding all control variables at their mean value. +Sample sizes for children gets very small when broken down by residence and caste; results should be treated with caution. +Predicted values differ slightly between Tables 3 and 4 because other variables are held at All-India mean in Table 3 while they are held at urban and rural specific means in Table 4 and full regional interactions are included allowing for differential impact of background factors by place of residence. NA Sample size < 50. SPECIAL ARTICLE comparison of caste differences across these four areas shown paints an interesting picture. On the whole, this table shows greatest inequality in developed villages and smaller cities, while the least developed villages and metro cities show lower levels of caste inequality. Although the results for the metro cities must be treated cautiously due to small sample sizes, particularly for adivasis, there are greater caste-based disparities in adjusted household expenditure than in other outcomes in metro cities. However, in other urban areas and developed villages, we continue to see substantial caste-based inequalities in all outcomes. How do we interpret these results? In some ways this pattern reminds us of the Kuznets inverted U-shaped curve which suggests that income inequality increases with development before declining at high levels of development (Kuznets 1963). In the present context, it seems likely that in relatively poor villages all residents suffer from low incomes but once incomes begin to rise as with developed villages or early urbanisation, the benefits accrue to privileged groups. It is only in metropolitan cities that castebased discrimination is moderated. We present these observations somewhat tentatively since this is one of the first studies to make such observation and our ability to generalise is limited by relatively small sample for metropolitan cities. However, our observations are consistent with some of the recent findings from relatively poor states (Kapur et al 2010) where caste inequalities have emerged as less salient than those in all-india studies based on larger samples and offer an explanation for these findings. 3.4 Caste as a Social Construction One of the most pernicious critiques of affirmative action rests on the argument that caste identities are socially constructed. When government policies rest on caste to define privilege, this very act creates and solidifies caste identities and has a counterproductive impact with underprivileged groups responding to surveys and censuses so as to highlight their marginalisation in order to garner government benefits. As with most social constructionist arguments, it is virtually impossible to examine the empirical validity of this argument. This paper is no exception. However, we are able to shed some light on at least a small portion of this argument. If caste identities are solidified around reservations or positive discrimination policies, we expect to see differences in reported measures of well-being between forward castes and groups subject to affirmative action namely, OBCs, dalits and adivasis. However, since forward castes are unaffected by this misreporting, there should be few differences between different jatis within forward castes. The IHDS is the only large national survey to differentiate within the forward caste category by asking respondents whether they are brahmins or belong to other forward castes. Tables 1-4 distinguish between these two groups on all outcomes studied. The results show strong and significant differences between brahmins and other forward castes on almost all outcomes. While brahmins form only 6% of the sample, they appear to be uniquely privileged. Brahmins are more likely to have high education, they are more likely to have higher incomes and consumption expenditure and greater social connections than other Economic & Political Weekly EPW march 12, 2011 vol xlvi no 11 45

7 SPECIAL ARTICLE forward castes. These differences are large and statistically significant in most of the dependent variables studied in Tables 2 and 3. Even for Table 4 where place of residence is strictly controlled through separate analyses and sample sizes for brahmins in each residential category become extremely small, brahmins continue to exhibit higher incomes and educational levels than other forward castes. This suggests that although politics continues to dominate identity formation in modern India, caste is not simply a social construction. Even within groups where caste mobilisation has not been as salient a phenomenon as is the case for the OBCs, we see substantial differences by caste. If this were not the case, we would not see continued dominance of brahmins on a variety of markers of social and economic well-being. 3.5 Caste and Social Stratification This paper has examined caste disparities in a variety of outcomes using data from a recent survey, the IHDS of The results paint an intriguing picture of caste stratification in modern India. On the one hand, there are promising signs of change as observed in high civic and political participation on the part of dalits and adivasis and lower levels of disparities in metro cities, on the other hand, it is impossible to claim that caste has ceased to define either the opportunity structure or outcomes for a vast proportion of Indian population. Results presented above suggest that caste background continues to define opportunities available to individuals. Landownership patterns remain unequal; lower castes have low educational status; have lower consumption expenditure resulting in lower access to nutrition, healthcare and private education; and have fewer social connections to seek help in emergencies or to provide access to information and connections to important social institutions such as government services, healthcare and medical services. This unequal opportunity and access may well be the root cause of observed inequalities in health outcomes as well as other aspects of well-being (IIPS and Macro International 2007; Desai et al 2010; Shariff 1999). However, if these material disadvantages are combined with an absence of social discrimination, then it would be easy to argue that caste has been transformed into class in modern India and public policies do not need to focus on caste policies directed at lower class individuals, regardless of their caste affiliation, should be sufficient. Figure 1: Predicted Total Income and Wage Income Before and After Controlling for Education Forward castes Brahmin OBCs Dalit Adivasi All incl unadjusted All incl adjusted Wage incl Wage incl for education unadjusted adjusted for education Have caste inequalities in modern India been transformed into class inequalities? If that were the case, then most of the caste disparities should disappear once controls for education and access to other productive resources are added. Our analysis shows no signs of that. We find that although educational disparities (sometimes called pre market inequalities ) account for about 50% of the intercaste disparities in earnings (Figure 1), even after controlling for own education, substantial caste disparities remain. Perhaps the most damning are the differences in skill acquisition of children. After controlling for highest level of education attained by any household member, log of per capita expenditure and place of residence, only about 50% dalit and adivasi children can read a simple paragraph compared to 61% forward class children and 69% brahmin children. The differences in reading skills between OBCs and forward castes are relatively small and not statistically significant. It is important to note that we are only focusing on rudimentary skills ability to read a simple 3-4 sentence paragraph. It would be difficult to justify young children aged 8-11 falling behind their upper caste peers on any other grounds but social forces that exclude some children and not others (for some of the examples of social exclusion see Navsarjan Trust and RFK Centre 2010). 4 Discussion 4.1 Large and Small Differences: The Eye of the Beholder In this paper we have focused on comparing like with like and consistently controlled for place and state of residence in our analysis. Where appropriate, we have also controlled for education and landownership. This is an appropriate strategy for examining differences associated purely with caste rather than place of residence and individual characteristics. However, these fine controls, a conventional strategy for statistical analysis in search of pure relationship between two variables of interest, sometimes get in the way of establishing the magnitude of the problem. When we look at Table 4, we see only modest differences between forwards castes and lower castes. For example, in least developed villages, after controlling for household education and landownership, the difference in household size adjusted consumption expenditure between forward castes and dalits is barely Rs 2,200, but when we look at raw differences between the two groups in Appendix 1, it is Rs 9,019. The raw differences show cumulative disparities based on a variety of factors including disproportionate concentration of dalits in the least developed villages (39% for forward castes vs 48% for dalits in Appendix 1), low likelihood of landownership (45% vs 34%), greater likelihood of no one the household ever attending school (10% vs 34%), lower social connections to obtain better jobs (1.33 vs.79), and lower wages for adult males when employed for salary or wage work (Rs 30,000 vs Rs 13,800). When regressions control for some of these factors which are themselves either closely associated with caste or are products of caste disadvantages, it is easy to understate caste-based inequalities. Spatial disparities are particularly interesting in this context. The IHDS data document that where one lives is associated with economic opportunities as well as availability of services such as march 12, 2011 vol xlvi no 11 EPW Economic & Political Weekly

8 healthcare and education (Desai et al 2010), sometimes these spatial disparities are even larger than caste disparities. However, spatial disparities are not totally orthogonal to caste disparities. As Appendix 1 shows, dalits and adivasis are far more likely to be concentrated in the least developed villages than forward castes, possibly because developmental interventions miss dalit and adivasi dominated areas. Moreover, migration is associated with social position. Most studies of migration in India and elsewhere document that often it is the more privileged groups in any society that migrate, possibly because higher education and social networks make it easier for them to find jobs in cities (NSSO 2001; Massey and Taylor 2004; de Haan and Rogly 2002). Among the IHDS sample, about 10-11% of the forward caste and brahmin households migrated in the 10 years prior to the survey compared to 7% for OBCs and less than 5% for dalits and adivasis. Moreover, while living in more developed areas offers higher incomes to all residents, caste disparities do not disappear with development. Our results show an inverted U-shaped relationship between spatial development and caste inequalities with least developed villages and metro cities showing lower caste inequalities than somewhat developed villages or smaller urban areas. Since much of the future growth seems likely to concentrate in smaller cities, this is a worrisome development. Due to our reliance on cross-sectional data, we cannot make any conjectures about changes over time but these spatial patterns are suggestive. This suggests that the public discourse must now begin to focus on ways in which economic growth and weakening of traditional occupational structures intersects with historically dominant caste divisions to restrict opportunities to some groups and opens opportunities to others particularly through access to high quality education. 4.2 Caste in the 21st Century: From Hierarchy to Elite Capture The results presented above show that if status hierarchies seem to be on the decline with considerable civic and political participation by marginalised groups, economic and educational disparities between large caste groupings continue to flourish. What are the implications of these observations for research on caste in 21st century India, as well as for the public policy? We suggest that for far too long, the discourse on caste in India has been governed by the vision of social hierarchy based on purity and pollution laid down by a colonial imagination. It may be time to look for new models of social stratification in India. One possibility would be to argue that caste relations are superseded by class relations, just as the western sociological literature has argued that industrialisation was associated with a movement from hereditary privilege associated with the concept of estate held by nobility to an open society in which the bourgeoisie could gain status through their achievement in the marketplace (Borocz 1997). However, the results presented above suggest that this would be a premature conclusion. Access to productive resources, particularly education and skills remain closely associated with caste. Children from lower castes continue to be educationally SPECIAL ARTICLE disadvantaged compared to children from the upper caste. Once stripped of its religious and ideological trope, caste in modern India offers one of the most interesting examples of consolidation of material resources in hands of certain groups even as market mechanisms continue to take hold. The continued dominance of brahmins in Indian society and economy is perhaps the clearest example of this consolidation. Newly emerging research on central and eastern Europe offers interesting comparisons. A number of studies suggest that the transition from socialism to capitalism in eastern and central Europe has led to the capture of economic resources by certain groups (Eyal et al 1998; Rona-Tas 1998). Given the explicit attempts by socialist states over a period of five decades to eradicate hereditary transmission of wealth and power, the persistence of privilege remains surprising. In reconciling scholarship on post-socialist transition with broader stratification research, Borocz (Borocz 1997) argues for a middle space between the societies organised around the logic of estate where the individual s position is determined by ascription based on membership in a group, conferred mainly by birth and logic of status in which the basis of the individual s position is exclusively achievement portrayed as returns on human capital endowments. His advocacy for a middle ground with a focus on social reproduction where new logics of distinction are created as enclosures in the social field, resonates with the observations offered in this paper. Instead of hard distinctions between caste and class (e g, Svalastoga 1965) based on varying degrees of closure within each form of stratification, we will gain more analytical power by focusing on ways in which a socially closed system like caste adapts and manipulates emerging class inequalities in a society undergoing economic transformation. The IHDS documents that caste and kin remain at the centre of Indian civic life, with nearly 95% of the female respondents reporting getting married within their own caste. This suggests that caste structure remains largely defined by social closure, while the economic opportunities are increasingly more open, creating new arenas in which castes must operate if they are to maintain their distinction. Seeing Indian caste structure through the lens of Bourdieu s notions of social reproduction (Bourdieu 1984), we begin to see a variety of ways in which castes manage to shape the access to social, political and cultural capital to their members. Much has been written about the political power of various castes (Gupta 2005) with a particular focus on the middle castes. However, the way in which upper castes manage to establish and exercise their dominance over the opportunities to their members has received little attention. As education and entrepreneurship emerge as twin pillars of advancement in modern India, historically wealthy castes play an interesting, often unnoticed, role in shaping opportunities. The IHDS data collected information on a variety of associations and finds that caste associations dominated organisational memberships with 14% households belonging to caste associations and another 14% belonging to religious or festival societies compared to barely 5% households with any union or trade association membership and 7% participating in self-help groups. Caste associations for wealthy castes organise private schools Economic & Political Weekly EPW march 12, 2011 vol xlvi no 11 47